Molecular Dynamics Simulation of the Microscopic Sintering Process

2 days ago - CuO-based materials as oxygen carrier (OC) always exhibit a weak sintering resistance at high temperature, which leads to a significant ...
1 downloads 0 Views 14MB Size
Article Cite This: J. Phys. Chem. C XXXX, XXX, XXX−XXX

pubs.acs.org/JPCC

Molecular Dynamics Simulation of the Microscopic Sintering Process of CuO Nanograins Inside an Oxygen Carrier Particle Haibo Zhao,* Jinfa Gui, Jie Cao, and Chaohe Zheng

Downloaded via KAOHSIUNG MEDICAL UNIV on October 27, 2018 at 18:25:30 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

State Key Laboratory of Coal Combustion, Huazhong University of Science and Technology, Wuhan, 430074, PR China ABSTRACT: CuO-based materials as oxygen carrier (OC) always exhibit a weak sintering resistance at high temperature, which leads to a significant decrease of reactivity in chemical looping processes. Inert component is usually added to enhance the thermal stability and increase the specific surface area of OC particles. Detailed knowledge on the sintering mechanism of CuO nanograins within the bulk of OC particles and the interactions between active component and inert support materials is thus of considerable importance. In this study, molecular dynamics (MD) method was conducted to explore the fundamental understanding of CuO sintering mechanism and the effects of different support materials (TiO2, ZrO2, and SiO2) on the sintering resistance of supported CuO nanograins. The sintering simulations of pure CuO nanograins show that CuO particle with smaller diameter or at higher temperature tends to be more amorphous. With respect to the sintering of two unsupported nanograins, it can be concluded that the neck growth during sintering is the joint effect of surface diffusion and grain boundary diffusion. Among these three composite OCs (CuO supported by TiO2, ZrO2, or SiO2), CuO/ZrO2 shows a better sintering resistance. The enlarged discrepancy on the surface area loss between different supported CuO nanograins with the rising of temperature emphasizes the importance of rational selection of support materials at high temperature.

1. INTRODUCTION It is well-believed that the CO2 emission from fossil fuel combustion is a major factor for the aggravation of greenhouse effect. The reduction of CO2 emission is in urgent need, and several CO2 capture and storage (CCS) technologies have been developed to mitigate this emission within power generation processes.1 Among these CCS technologies chemical-looping combustion (CLC) is a type of low-cost CO2 capture technology and considered to be a promising lowcarbon combustion/utilization technology of carbon-intensive fossil fuels.2 CLC faces a tough challenge when utilizing solid fuels (e.g., coal) as the char gasification reaction is a ratelimiting step. A variant CLC process called chemical looping with oxygen uncoupling (CLOU)3 was proposed to solve this problem. The key of CLOU is to identify high-performance oxygen carriers (OC) that are able to release gaseous O2 at high temperature (800−1000 °C) and oxygen-deficient atmospheres. As O2 is generated in fuel reactor (FR) during CLOU processes, it will help accelerate the char conversion, which eventually contributes to much higher CO2 capture efficiency and combustion efficiency. CuO is a superior candidate of CLOU oxygen carrier due to its high reactivity, high oxygen transport capacity, medium price, moderate environment impact, being exothermic for fuel combustion in FR, and suitable equilibrium partial pressure of oxygen at temperature of interest for combustion (800−1000 °C).4,5 Despite of these favorable material properties, CuO with a quite low Tammann temperature (526 °C) suffers from a tendency toward sintering at CLOU temperature, which will lead to defluidization and low reactivity of the Cu-based OC particles. Usually, CuO is supported by inert materials to © XXXX American Chemical Society

withstand high reaction temperatures. To date, a lot of materials including TiO2, ZrO2, SiO2, Al2O3, MgAl2O4, and CuAl2O44,6,7 have been used as the support for CuO due to their thermally stability and high melting point. Various method (e.g., impregnation, sol−gel, mechanical mixing, spray drying) are adopted for the preparation of supported-CuO OC particles, and the preparation processes are almost involved in drying, granulation, calcination, grinding, and sieving.8−10 How to choose appropriate support materials to enhance the sintering resistance becomes a challenge for the preparation of Cu-based OC. The identification of high-performance OC based on the trial-and-error experiments is costly and time-consuming.4,11 The Zener equation,12 which is based on the idea that the driving pressure for first-phase grain growth (accompanying with particle sintering) due to the curvature of the grain boundary would be counteracted by a pinning pressure exerted by the second-phase particles (support materials) on the boundary, can be used to describe the active-component grain growth phenomenon as a function of the support material fraction, the type and size of the material, therefore helping rationalize the selection of inert materials.13,14 However, the Zener pinning theory cannot provide any information about the actual process of grain growth, the growth rate or the size distribution of grains as a function of time,15 let alone the morphology evolution and motion trail of particles during the high-temperature sintering processes. An intuitionistic guidReceived: May 4, 2018 Revised: October 22, 2018 Published: October 22, 2018 A

DOI: 10.1021/acs.jpcc.8b04253 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

condition was adopted, and the cutoff radius was set as 0.95 nm, which was also adopted by some references.24,25 The cutoff length is larger than the default Cutoff_Element values for “Fine” quality calculations (for O element, the default cutoff value for Fine quality calculations is 0.33 nm; 0.44 nm for Cu; 0.52 nm for Ti; 0.46 nm for Si; and 0.53 nm for Zr; see Materials Studio online help). To enable the canonical ensemble simulations, the Nose−Hoover temperature coupling26 was selected to relax the atoms in the sintering system. Since the objective metal oxides in this study are various (CuO, TiO2, ZrO2, SiO2), the force fields that are suitable for all these metal oxides are very limited. The COMPASS (condensed-phase optimized molecular potential for atomistic simulation studies) force field,27 which is able to describe the interaction of many metal oxides, is an ideal choice for the calculations. In the COMPASS force field, a semi-ionic model27 was used. Some involved parameters and physical properties of the COMPASS force field for the simulation of CuO,28 TiO228, and SiO228−30 have been fully described in these references and well parametrized in Materials Studio. To the best of our knowledge, there is no such parameter description of ZrO2 for the COMPASS force field from Materials Studio online help or any open reference. Materials Studio online help recommends that the semi-ionic model can be used for many metal oxides. In this work, we then used the COMPASS force field to simulate the ZrO2 system. Factually, we only use the relevant simulation parameters and targeted physical properties in the Discover package of Materials Studio software. We did not define or modify these independence parameters and targeted physical properties in the MD simulations. Atomic interactions in this force field model are divided into two types: directly contacted (bonded) interaction and indirectly contacted interaction. For the atoms that are directly bonded, the interaction energy is described by the Morse potential function and an electrostatic term:28

ance on rational selection of inert supports may be provided by atomic-scale analyses about the interaction between active component and support, and a well understanding of the microcosmic sintering process is also of great benefit to the preparation of OC particles, including the optimization of particle formulation and dispersion. Essentially, the sintering behavior of OC is dependent upon the microstructure, from the grain size to lattice arrangement. We16 have used the density functional theory (DFT) to analyze the effect of support on the sintering resistance of supported-CuO particles, based on the analyses of adsorption energy and bond lengths. But it is still a nonintuitionistic criterion for the sintering resistance. Obviously, the sintering resistance is also associated with, at least, the grain size of active and inert components, the dispersion mode and mixing ratio between two components, which is hard to be handled by DFT since it is computationally infeasible to simulate the sintering process of nanograin system with too many atoms involved. Molecular dynamics (MD) simulation represents a powerful tool that can provide insights on the molecular mechanisms that drive and control the interactions at nanoscale. By solving the equation of motion for a system of particles, the real behavior of materials can be simulated under a specified temperature. The trajectory and physical movement of atoms, molecules, and nanoparticles (NPs) in the system can also be determined. MD simulations have been successfully used to investigate the sintering performance of metal oxide at high temperature.17−21 Schweigert et al.17 studied the properties of SiO2 clusters at temperatures from 1500 to 2800 K, and found that the diffusion coefficient plays a critical role in determining the primary particle size. Buesser et al.19 studied the sintering of small (20 to 40 Å) rutile TiO2 nanoparticles to complete coalescence, which showed that lower temperatures or larger primary particle diameters leads to slower sintering. Ahmed et al.21 reported a detailed analysis of the nanoscale structure of CuO NPs as functions of size and temperature using MD simulations. It was found that the melting temperature is inversely proportional to the particle diameter, and the surface exhibits a distorted outer shell below the melting point. However, this study only paid attention to the melting feature of single CuO nanoparticle. In a realistic process, the sintering between different particles may lead to a more significant reduction of specific surface area, and the interaction between CuO and refractory material can inhibit the sintering effectively. Although some publications have investigated the sintering properties of composite particles (e.g., nickel/zirconia system22,23), to the best of our knowledge, the MD simulation on supported-CuO material has not been reported yet. In this paper, a detailed description of the sintering characteristics of CuO nanograins (NGs) was investigated via MD simulations. The calculations started with the exploration of pure CuO nanograin sintering mechanism, where the effects of temperature and particle size were studied. Then, a cluster-slab model was used to clarify the role of different support materials in determining the sintering resistance of CuO at different temperatures.

Eij = ε{exp[−2α(rij − rij0)] − 2 exp[−α(rij − rij0)]} × fs + (1 − fs )

C6 rij 6

+

keqiqj rij

(1)

r0ij

where rij is the interatomic distance in Å, is the equilibrium bond distance in Å, ε is trap depth energy in eV, C6 is a parameter related to the bond energy, ke is Coulomb’s constant, the “width” of the potential is controlled by α in Å−1, q is the partial charge of atom in e, and fs is a continuous switching factor, which is defined as fs =

1 1 + exp[20(rij − rc)]

(2)

where rc is the radiation wavelength in Å. For the interactions between indirectly contacted atoms, the energy is described by the electrostatic term and VDW (van der Waals) term: É ÅÄÅ i 6 Ñ ÅÅ ji r 0 y9 ij rij0 yz yzzÑÑÑÑ keqiqj ÅÅ jjjj ij zz z Eij = εÅÅÅ2jjjjjj zzz − 3jjjj zzzz zzzÑÑÑÑ + j rij z zzÑÑ ÅÅ jjj rij z rij ÅÅ kk { k { {ÑÑÑÖ (3) ÅÇ

2. COMPUTATIONAL DETAILS The MD simulations were performed by the Discover package in Materials Studio. The constant-temperature, constantvolume canonical ensemble (NVT) was applied in the sintering simulations with a time step of 1.0 fs. The periodic boundary

In this simulation, the “Ewald method” was adopted for the charge term28,31 keqiqj/rij of COMMPAS potential. On the basis of the help files of the Discover package in Materials Studio, we selected the summation method of “atom based” to B

DOI: 10.1021/acs.jpcc.8b04253 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C describe the van der Waals interactions, and adopted the Ewald method to evaluate Coulomb interactions with parameters of medium quality. With respect to CuO (as an example), there are three types of atoms interaction: Cu−Cu, O−O, and Cu−O interactions. The interaction is divided into a bonded form and a nonbonded form, but only Cu−O bond interaction is formed in the bonded form, and the Morse potential function interaction is shown as Figure 1.

Figure 3. Coulombic potential.

In order to guarantee the computational accuracy of the MD simulation, the force field method and simulation parameters were tested. We first performed the geometry optimization calculation of CuO unit cell with these parameters described above, and the calculation results of Cu−O bond length at 0 K were given in Table 1. The MD result is close to the Table 1. Cu−O Bond Length in the Optimized CuO Unit Cell Figure 1. Morse potential.

The interaction potential exhibits a steady state of lowest energy, and it rises sharply as the bond distance decreases, then tends to zero as the bond distance becomes longer. For nonbonded interactions, as shown in Figure 2 and 3, the potential functions are divided into Lennard-Jones potential and Coulomb potential. Comparatively speaking, the interaction of the bonded form is stronger than that of the nonbonded form. The nonbond interaction is very weak, but in its effective intercept radius space, the number of atoms that undergo nonbonding interactions is large and is applied to the atoms in superimposed form.

method

MD

DFT32

experiment33

length of Cu−O bond (Å)

1.98

1.95

1.96

experimental value33 as well as the DFT result.32 The high accuracy of Cu−O bond length is sufficient for the MD simulation of large amount of CuO molecules. Then, we calculated the lattice parameters of the three inert supports of TiO2, ZrO2 and SiO2 and compared then with the experimental data. As summarized in Table 2, the calculated results are very close to the experimental values,34−36 indicating the validity of parameters setting in this work. CuO is a monoclinic crystal and belongs to the C2/C1 space group. In this study, the CuO nanograins were assumed to be ideal spheres, and the initial CuO NGs with different diameters (30, 40, and 50 Å) in the sintering simulation were built by a spherical cut from the CuO super lattice, and excess atoms were removed from the surface. The constructed CuO cluster with diameter of 30 Å contains 1376 atoms, 3344 atoms for cluster diameter of 40 Å, and 6552 atoms for cluster diameter of 50 Å. Figure 4 presents the initial module of CuO cluster of 40 Å. The stable low-index surface of support materials, TiO2 (110), ZrO2 (100), and SiO2 (110), were selected as the calculated surface in this work, which were also cleaved from the bulk phase of TiO2, ZrO2 and SiO2. The size of the TiO2 (110), ZrO2 (100), and SiO2 (110) surface are 100 Å × 160 Å. The module of CuO cluster loaded on the surface of support material TiO2 can be seen in Figure 5. In this work, the support materials of TiO2, SiO2 and ZrO2 have been fixed, which was also similar to the simulation condition of ref 22. The module of CuO nanograins loaded on the fixed support surface (Figure 5) can be viewed as a simplification of a real oxygen carrier particle prepared by the impregnation method,

Figure 2. Lennard-Jones potential. C

DOI: 10.1021/acs.jpcc.8b04253 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C Table 2. Lattice parameters of TiO2, ZrO2 and SiO2 support (group space)

calculation (this work)

experiment

TiO2 (P42/MNM) ZrO2 (P21/C) SiO2 (P3121)

a = b = 4.48 Å; c = 3.00 Å; α = β = γ = 90° a = 5.17 Å; b = 5.11 Å; c = 5.32 Å; α = γ = 90°; β = 99° a = b = 4.89 Å; c = 5.36 Å; α = β = 90°; γ = 120°

a = b = 4.58 Å; c = 2.95 Å; α = β = γ = 90°34 a = 5.15 Å; b = 5.21 Å; c = 5.32 Å; α = γ = 90°; β = 99°35 a = b = 4.91 Å; c = 5.41 Å; α = β = 90°; γ = 120°36

support molecules (atoms) will be quite gentle in the MD simulations, and we then ignored the internal movement of support molecules (atoms). In this work, the thicknesses of support materials, about 2 nm (nine-layer structure), referred to Xu et al.’s setting for simulating the sintering process of Ni nanoparticles in Ni/YSZ multinanoparticle system.22 Because the support materials have been fixed, we therefore speculated that the thicknesses of support materials (about 2 nm) are feasible for modeling of bulky-particles. The residual stress and strain were not considered in the MD simulations. We also noted here that the thicknesses of support materials (ca. 2 nm) are larger than the cutoff length (0.95 nm).

Figure 4. CuO cluster initial model.

3. RESULTS AND DISCUSSION 3.1. High-Temperature Sintering Characteristics of Pure CuO Grain. The structural properties of single CuO NG belong to vital and fundamental information to understand the CuO sintering behavior. A pure CuO particle with the diameter of 40 Å was put into a large cubic box (the length of side is 100 Å) and heated to the target temperature by Nose-Hoover temperature coupling for a total time of 1 ns. After this process, the radial distribution function (RDF) of CuO NG was counted to examine the structure features. The result is shown in Figure 6a. Figure 6a shows that the RDF exhibits distinct characteristic peaks at 973, 1173, and 1373 K, and these significant peaks were all observed in the RDF curves at three different temperatures, which indicates that the CuO particles in high temperatures will still maintain a relative regular crystal lattice arrangement. The first strongest peak appears in the distance of 1.99 Å for all the three temperatures, corresponding to the Cu−O bond length in central position, which is very close to the ideal bond length of CuO crystal (1.95 Å). In addition, it also can be found that the RDF curves at higher temperatures tend to be slightly flatter, indicating the high-temperature

Figure 5. Atomic structure of CuO NGs and the support surface.

in which a small amount of active components (here CuO nanograins, usually with loading of